Study on TLS Point Cloud Registration Algorithm for Large-Scale Outdoor Weak Geometric Features

Abstract

With the development of societies, the exploitation of mountains and forests is increasing to meet the needs of tourism, mineral resources, and environmental protection. The point cloud registration, 3D modeling, and deformation monitoring that are involved in surveying large scenes in the field have become a research focus for many scholars. At present, there are two major problems with outdoor terrestrial laser scanning (TLS) point cloud registration. First, compared with strong geometric conditions with obvious angle changes or symmetric structures, such as houses and roads, which are commonly found in cities and villages, outdoor TLS point cloud registration mostly collects data on weak geometric conditions with rough surfaces and irregular shapes, such as mountains, rocks, and forests. This makes the algorithm that set the geometric features as the main registration parameter invalid with uncontrollable alignment errors. Second, outdoor TLS point cloud registration is often characterized by its large scanning range of a single station and enormous point cloud data, which reduce the efficiency of point cloud registration. To address the above problems, we used the NARF + SIFT algorithm in this paper to extract key points with stronger expression, expanded the use of multi-view convolutional neural networks (MVCNN) in point cloud registration, and adopted GPU to accelerate the matrix calculation. The experimental results have demonstrated that this method has greatly improved registration efficiency while ensuring registration accuracy in the registration of point cloud data with weak geometric features.

Automated Point Cloud Registration Approach Optimized for a Stop-and-Go Scanning System

“…The point cloud data acquired by TLS at an extensive indoor site exhibits the following characteristics: Firstly, their density is irregular, with points closer to the sensor being densely distributed, while those farther away are sparsely represented. Secondly, the data volume is substantial, with a single station’s TLS data comprising several million points in sizable scenarios, demanding significant computational resources [ 48 ]. To address this issue, several studies have adopted voxel-based filtering for point cloud downsampling [ 28 , 42 , 48 , 69 ] because the subsampled point cloud has a more even point density and lower point count, which benefits the efficiency of registration [ 28 , 69 ].…”

“…Secondly, the data volume is substantial, with a single station’s TLS data comprising several million points in sizable scenarios, demanding significant computational resources [ 48 ]. To address this issue, several studies have adopted voxel-based filtering for point cloud downsampling [ 28 , 42 , 48 , 69 ] because the subsampled point cloud has a more even point density and lower point count, which benefits the efficiency of registration [ 28 , 69 ]. Voxel filtering is used to preprocess the point cloud, reducing the influence of noise points on registration accuracy and improving efficiency.…”

“…Ao et al [ 47 ] presented a deep-learning-based surface descriptor called SpinNet for 3D point cloud registration. Li et al [ 48 ] designed the point cloud multi-view convolution neural network (PC-MVCNN) for point cloud registration. However, deep-learning-based approaches face a critical limitation due to their dependence on the training dataset.…”